Efficient Spoofing Attack Detection against Unknown Sample using End-to-End Anomaly Detection
Abstract
With the evolution of a high precision sensor, printing machine, and manufacturing machine, spoofing attacks become a significant threat to the biometric systems. In order to mitigate the threats of diverse and unexpected attacks, conventional spoofing attack detection methods which aim to detect a specific attack are not sufficient. In this study, we propose a end-to-end machine learning technique which can model biometric information with the complicated structure of high dimension by a probability distribution. The proposed system can recognize whether inputted sample is spoofing one or not even if it is an unknown attack.